Comparison of two home sleep testing devices with different strategies for diagnosis of OSA

Home sleep testing devices are being widely used in diagnosis/screening for obstructive sleep apnea (OSA). We examined differences in OSA metrics obtained from two devices with divergent home monitoring strategies, the Apnea Risk Evaluation System (ARES™, multiple signals plus forehead reflectance oximetry) and the Nonin WristOx₂™ (single channel finger transmission pulse oximeter), compared to differences from night-night variability of OSA.

One hundred fifty-two male/26 female subjects (BMI = 30.3 ± 5.6 kg/m², age = 52.5 ± 8.9 years) were recruited without regard to OSA symptoms and simultaneously wore both ARES™ and Nonin WristOx₂™ for two nights (n = 351 nights). Automated analysis of the WristOx₂ yielded oxygen desaturation index (ODI_Ox2, ≥4% O2 dips/h), and automated analysis with manual editing of ARES™ yielded AHI4_ARES (apneas + hypopneas with ≥4% O2 dips/h) and RDI_ARES (apneas + hypopneas with ≥4% O2 dips/h or arousal surrogates). Baseline awake oxygen saturation, percent time < 90% O2 saturation (%time < 90%O2Sat), and O2 signal loss were compared between the two methods.

Correlation between AHI4_ARES and ODI_Ox2 was high (ICC = 0.9, 95% CI = 0.87-0.92, p < 0.001, bias ± SD = 0.7 ± 6.1 events/h). Agreement values for OSA diagnosis (77-85%) between devices were similar to those seen from night-to-night variability of OSA using a single device. Awake baseline O2 saturation was significantly higher in the ARES™ (96.2 ± 1.6%) than WristOx₂™ (92.2 ± 2.1%, p < 0.01). There was a significantly lower %time < 90%O2Sat reported by the ARES™ compared to WristOx₂ (median (IQR) 0.5 (0.0, 2.6) vs. 2.1 (0.3, 9.7), p < 0.001), and the correlation was low (ICC = 0.2).

OSA severity metrics predominantly dependent on change in oxygen saturation and metrics used in diagnosis of OSA (AHI4 and ODI) correlated well across devices tested. However, differences in cumulative oxygen desaturation measures (i.e., %time < 90%O2Sat) between the devices suggest that caution is needed when interpreting this metric particularly in populations likely to have significant hypoxia.